Re: two ideas...

[email protected]
Wed, 29 Nov 1995 14:26:39 -0800


> I have two ideas which I would like to get some feedback on.
> [...] The second is to implement intelligent pre-fetching for the web.
>
> You might be interested in looking at Venkata Padmanabhan's MS
> thesis from UC Berkeley, available as
> http://www.cs.berkeley.edu/~padmanab/papers/masters-tr.ps
> Venkat used a trace-based simulation to investigate the potential
> performance of prefetching. Some of this work was done under my
> supervision.
>
> Apparently some people at Boston University have done similar work,
> since their CS technical Reports page
> http://cs-www.bu.edu/techreports/Home.html
> lists
> 95-011
> Azer Bestavros ; Carlos Cunha.
> A Prefetching Protocol Using Client Speculation for the WWW,
> May 8, 1995.
> However, clicking on the link to this report yields: "Error 403
> Forbidden - by rule" and I have not been able to obtain a copy.
>
> -Jeff

We've also been looking at Web prefetching, using server-side
speculation.

References are listed on our Web pages:

http://www.isi.edu/lowlat/

Notably
"Defining 'High Speed Protocols, Five Challenges & and Example
That Survives the Challenges," J. Touch, IEEE JSAC Special
Issue on Applications Enabling Gigabit Networks, Vol. 13, No.
5, June 1995, pp. 828-835.

"An Experiment in Latency Reduction,", J. Touch and D. Farber
IEEE Infocom 1994.

We found that server speculation would decrease latency by 2/3, to 0.7
RTT (yes, below the speed of light) by increasing the BW by 7x. Note
that this RTT is an average per page - it still takes 1 RTT for the
first page...

Joe
----------------------------------------------------------------------
Joe Touch [email protected]
ISI / Project Leader, ATOMIC-2 http://www.isi.edu/~touch
USC / Research Assistant Prof. http://www.isi.edu/atomic2